An adjustable resource response capability probabilistic characterization and evaluation method

Abstract

The application provides a method for probabilistic characterization and evaluation of adjustable resource response capability, the characterization method comprising constructing a response capability model of each adjustable resource and a probability distribution model based on response capability influencing factors; using a sampling method, obtaining an influencing factor scenario set of a multi-resource power system based on the probability distribution model, and solving the response capability model to obtain an initial sample scenario set and a response capability index value set within a scheduling period; clustering the initial sample scenario set to obtain a response capability sample scenario set to obtain a scenario probability set and a probabilistic response capability index of each scheduling period within the scheduling period. The method can realize unified characterization of different types of resource response capability, quantitatively describe the randomness, volatility and time-varying characteristics of adjustable resource response capability, and provide a basis for subsequent resource aggregation evaluation, flexibility quantification and scheduling decision.

Description

Technical Field

[0001] This application relates to the field of power resource response capability assessment technology, specifically to a probabilistic characterization and assessment method for adjustable resource response capability. Background Technology

[0002] With the development of new power systems, the scale of adjustable resources such as electric vehicles, air conditioning loads, data centers, 5G base stations, and distributed energy storage is constantly expanding. These resources are characterized by wide distribution, diverse types, and rapid adjustment speed, playing a crucial role in demand response, peak shaving and valley filling, and flexible regulation. Therefore, accurately characterizing the response capabilities of various types of adjustable resources has become a key issue in resource aggregation assessment and resource regulation decision-making.

[0003] Existing methods for assessing the responsiveness of adjustable resources mostly employ deterministic parameters, typically providing only the responsiveness boundary of a resource under ideal conditions or given parameter conditions, and then performing a static analysis of its adaptability. While these methods can characterize the basic responsiveness of a resource to some extent, they fail to accurately reflect the uncertainties inherent in the resource's actual operation.

[0004] Furthermore, existing technologies still have significant shortcomings in handling uncertainty. On the one hand, most related studies focus on analyzing single types of resources or single random factors, making it difficult to achieve a unified characterization of the response capabilities of different types of resources. On the other hand, although some existing methods introduce random sampling or scenario analysis, they usually directly process resource response power or load curves, making it difficult to accurately reflect the process by which changes in influencing factors affect resource response capabilities. Summary of the Invention

[0005] This application addresses the problems existing in the prior art by providing a probabilistic characterization method for adjustable resource response capabilities. This method can achieve a unified characterization of the response capabilities of different types of resources, quantitatively describe the randomness, volatility, and time-varying characteristics of adjustable resource response capabilities, and provide a foundation for subsequent resource aggregation assessment, flexibility quantification, and scheduling decisions.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, embodiments of this application propose a probabilistic characterization method for the response capability of adjustable resources, applied to a multi-resource power system, which includes at least two adjustable resources. The method includes: constructing a response capability model for each adjustable resource; and obtaining the influencing factors of the response capability of each adjustable resource. Based on the factors affecting responsiveness, a probability distribution model is constructed for each adjustable resource. A sampling method was used to obtain a set of influencing factors for multi-resource power systems based on a probability distribution model; Based on the scenario set of influencing factors, the initial sample scenario set and response capability index value set within the scheduling period are obtained by solving the response capability model. Clustering is performed on the initial sample scene set to obtain a response capability sample scene set; A scenario probability set is obtained based on the initial sample scenario set and the response capability sample scenario set; Based on the scenario probability set, the response capability sample scenario set, and the response capability index value set, the probabilistic response capability index is obtained for each scheduling period within the scheduling cycle.

[0007] In some embodiments, K-means clustering is used to cluster the initial sample scene set to obtain the response capability sample scene set.

[0008] In some embodiments, K-means clustering is used to cluster the initial sample scene set to obtain a response capability sample scene set, including: The elbow rule is used to obtain the total number of clusters. The sum of squared errors of the initial sample scene set is used as the error function; Based on the total number of clusters and the error function, the K-means clustering method is used to cluster the initial sample scene set to obtain the response capability sample scene set.

[0009] In some embodiments, based on the total number of clusters and the error function, K-means clustering is used to cluster the initial sample scene set to obtain a response capability sample scene set, including: A total of 10 initial cluster sets are extracted from the initial sample scenario set; The initial cluster set is used as the current cluster set, and the geometric center of each sample in the initial cluster set is used as the current cluster center. Calculate the distance from each initial sample scene in the initial sample scene set to each current cluster center; The initial sample scene set is divided into a total of clusters based on the shortest distance, and the cluster sets are updated accordingly. Determine whether the error function has converged based on the updated cluster set and the current cluster set; If the error function does not converge, the updated cluster set will be used as the current cluster set, the geometric center of each sample in the updated cluster set will be used as the current cluster center, and the distance from each initial sample scene in the initial sample scene set to each current cluster center will be calculated repeatedly. If the error function converges, the cluster set will be updated as the response capability sample scenario set.

[0010] In some of these embodiments, at least two adjustable resources include electric vehicles, air conditioning loads, data centers, 5G base stations, and distributed energy storage; The response capability models include electric vehicle response capability model, air conditioning load response capability model, data center response capability model, 5G base station response capability model, and distributed energy storage response capability model. The electric vehicle response capability model is used to characterize the changes in charging power, discharging power and remaining power of electric vehicles during each scheduling period within the scheduling cycle. The air conditioning load response capability model is used to characterize the changes in the average operating power of air conditioners during each scheduling period within the scheduling cycle. The data center response capability model is used to characterize the adjustment of server computing power consumption and cooling system power consumption during each scheduling period within the scheduling cycle under different load scheduling strategies. The 5G base station response capability model is used to characterize the dynamic power adjustment amount and the dispatchable power of backup energy storage of the 5G base station in each scheduling period within the scheduling cycle, as well as the changes in the charging power, discharging power and remaining power of the energy storage in the 5G base station in each scheduling period within the scheduling cycle. The distributed energy storage response capability model is used to characterize the relationship between the charging power, discharging power and remaining power of distributed energy storage during each scheduling period within the scheduling cycle.

[0011] In some embodiments, the factors affecting response capability include those affecting electric vehicle response capability, those affecting air conditioning load response capability, those affecting data center response capability, those affecting 5G base station response capability, and those affecting distributed energy storage response capability. Factors affecting the responsiveness of electric vehicles include the time of electric vehicle registration with the grid and the time of electric vehicle deregistration. Factors affecting air conditioning load response capability include whether the user turns on the air conditioner, the indoor temperature set by the user, and the air conditioner temperature hysteresis control range. Factors affecting data center responsiveness include workload; Factors affecting the responsiveness of 5G base stations include user behavior factors; Factors affecting the response capability of distributed energy storage include users' willingness to accept regulation.

[0012] In some embodiments, the probability distribution models include electric vehicle probability distribution models, air conditioning load probability distribution models, data center probability distribution models, 5G base station probability distribution models, and distributed energy storage probability distribution models. A Gaussian mixture model is used to model the joint probability distribution of factors affecting the response capability of electric vehicles to obtain a probability distribution model of electric vehicles; The air conditioning load probability distribution model is constructed based on the factors affecting the air conditioning load response capability; A Gaussian mixture model was used to model the factors affecting the response capability of data centers to obtain a probability distribution model of data centers; A probability distribution model of 5G base stations was obtained by modeling the influencing factors of 5G base station response capability using the Gaussian Copula function. A probability distribution model for distributed energy storage was constructed based on the factors influencing the response capability of distributed energy storage using a logistic regression model.

[0013] In some embodiments, the Latin hypercube sampling method is used to obtain the set of influencing factors of multi-resource power systems based on a probability distribution model.

[0014] In some embodiments, Latin hypercube sampling is used to obtain a set of influencing factors for multi-resource power systems based on a probability distribution model, including: Obtain the initial total number of samples, which represents the total number of initial sample scenarios in the initial sample scenario set; The corresponding cumulative probability distribution function is obtained for each probability distribution model; Based on the initial total number of samples, the value range of each cumulative probability distribution function is evenly divided to obtain a set of value sub-intervals that correspond one-to-one with the cumulative probability distribution function; For each sub-interval of values, obtain the initial sample value; Substitute the initial sampled values ​​into the inverse function of the corresponding cumulative probability distribution function to obtain the sampled values ​​of the corresponding factors affecting response capability; Based on the probability distribution model, the sampled values ​​corresponding to each value sub-interval are arranged and combined to obtain the influencing factor scenario set.

[0015] Secondly, this application proposes an adjustable resource response capability assessment method, which uses a probabilistic response capability index to assess the adjustable resource response capability of a multi-resource power system. The probabilistic response capability index is obtained using a probabilistic characterization method for adjustable resource response capability.

[0016] Compared with the prior art, this application has the following advantages: 1. This application aims to characterize the sources of uncertainty in the response capabilities of different types of adjustable resources by identifying key factors affecting the response capabilities of various types of schedulable resources, namely, response capability influencing factors. Based on the statistical characteristics of these influencing factors, a corresponding probability distribution model is established to accurately reflect the effect of changes in these factors on resource response capabilities. This model serves as input for subsequent scenario generation and probabilistic representation of response capabilities. The response capability sample scenario set obtained through clustering and the probability of occurrence of each response capability sample scenario are uniformly quantified to form a probabilistic representation of the response capabilities of adjustable resources. This allows for a unified characterization of the response capabilities of different types of resources. The probabilistic representation results are used to quantitatively describe the randomness, volatility, and time-varying characteristics of the response capabilities of adjustable resources.

[0017] 2. This application, based on consideration of resource physical operation constraints and user behavior characteristics, first establishes response capability models for different types of resources; then identifies key random factors affecting resource response capability and establishes corresponding probability distribution models; further extracts typical response capability scenarios through scenario generation and scenario reduction methods; and finally forms probabilistic representation results of adjustable resource response capability, so as to achieve a quantitative description of the randomness, volatility, and time-varying characteristics of the response capability of multiple types of adjustable resources, thereby providing a foundation for subsequent resource aggregation assessment, flexibility quantification, and scheduling decisions. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the probabilistic characterization method for adjustable resource response capability in the embodiments of this application. Figure 2 This is a schematic diagram of the process in this application embodiment of using K-means clustering to cluster the initial sample scene set based on the total number of clusters and the error function to obtain the response capability sample scene set; Figure 3 This is a schematic diagram illustrating the process of obtaining the influencing factor scenario set of a multi-resource power system using the Latin hypercube sampling method based on a probability distribution model in the embodiments of this application. Detailed Implementation

[0019] To address the problems in the background technology, existing methods for assessing the responsiveness of adjustable resources based on deterministic parameter descriptions can characterize the basic responsiveness of resources to a certain extent, but they do not adequately consider random factors such as user behavior, business load, environmental conditions and operating conditions, making it difficult to truly reflect the uncertainty characteristics of resources in actual operation.

[0020] In handling uncertainty, analyses focusing on a single type of resource or a single random factor lack a unified probabilistic representation method for the response capabilities of multiple types of adjustable resources, making it difficult to achieve a unified characterization of the response capabilities of different resource types. Analytical methods that introduce random sampling or directly process resource response power or load curves fail to establish probabilistic models based on the key factors affecting resource response capabilities, thus making it difficult to accurately reflect the impact of changes in influencing factors on resource response capabilities.

[0021] To clearly illustrate the technical features of this solution, the implementation methods of this application will be described in detail below with reference to the accompanying drawings and embodiments. This will allow for a full understanding and implementation of how this application uses technical means to solve technical problems and achieve corresponding technical effects. The embodiments of this application and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this application.

[0022] See Figure 1In a first aspect, embodiments of this application propose a probabilistic characterization method for the response capability of adjustable resources, applied to a multi-resource power system, which includes at least two adjustable resources. The method includes: constructing a response capability model for each type of adjustable resource; obtaining the response capability influencing factors for each type of adjustable resource; identifying key factors affecting the response capability of various types of adjustable resources, i.e., response capability influencing factors, to characterize the sources of uncertainty in the response capability of different types of adjustable resources; and establishing a corresponding probability distribution model based on the statistical characteristics of the response capability influencing factors, as input for subsequent scenario generation and probabilistic characterization of response capability. Based on the factors affecting responsiveness, a probability distribution model is constructed for each adjustable resource. A sampling method was used to obtain a set of influencing factors for multi-resource power systems based on a probability distribution model; The response capability model is solved based on the influencing factor scenario set to obtain the initial sample scenario set and response capability index value set within the scheduling period. The response capability index value set is the set of response capability index values ​​for each scheduling period within the scheduling period. Clustering is performed on the initial sample scene set to obtain a response capability sample scene set, thereby reducing the scene size and retaining the main statistical features; A scenario probability set is obtained based on the initial sample scenario set and the response capability sample scenario set; the scenario probability set is: ; in, For the first The probability of scenarios in which sample scenarios of similar responsiveness appear. The response capability sample scenario is numbered in the response capability sample scenario set. For the first The number of scenarios in the sample scenarios of class response capability This represents the total number of initial sample scenes in the initial sample scene set. The total number of response capability sample scenarios in the sample scenario set; Based on the scenario probability set, the response capability sample scenario set, and the response capability index value set, the probabilistic response capability index is obtained for each scheduling period within the scheduling cycle. The probabilistic response capability index is: ; in, For the first Probabilistic response capability metrics for each scheduling period The scheduling time period number, For the first Class of response capability sample scenarios in the first The response capability index value for each scheduling period; To further characterize the overall response potential of adjustable resources within the scheduling cycle, a comprehensive response capability index is obtained based on the probabilistic response capability index: ; in, To provide comprehensive response capability indicators, This represents the total number of scheduling periods within the scheduling cycle. For the first The weighting coefficient for each scheduling period. For the first The duration of each scheduling period.

[0023] Beneficially, this application identifies key factors affecting the response capabilities of various types of schedulable resources, namely, response capability influencing factors, to characterize the sources of uncertainty in the response capabilities of different types of schedulable resources. Based on the statistical characteristics of the response capability influencing factors, a corresponding probability distribution model is established as the input for subsequent scenario generation and probabilistic representation of response capabilities. The response capability sample scenario set obtained through clustering and the probability of occurrence of each response capability sample scenario are uniformly quantified to form a probabilistic representation result of the responsive capability of schedulable resources. The probabilistic representation result includes scenario probability, probabilistic response capability index, and comprehensive response capability index, which are used to quantitatively describe the randomness, volatility, and time-varying characteristics of the responsive capability of schedulable resources.

[0024] In some of these embodiments, at least two adjustable resources include electric vehicles, air conditioning loads, data centers, 5G base stations, and distributed energy storage; The response capability models include electric vehicle response capability model, air conditioning load response capability model, data center response capability model, 5G base station response capability model, and distributed energy storage response capability model. The electric vehicle (EV) response capability model characterizes the changes in charging power, discharging power, and remaining energy of an EV during each scheduling period within a scheduling cycle. The EV's response capability is primarily constrained by its state of charge (SBC), upper limits of charging and discharging power, charging and discharging efficiency, and grid connection and off-grid time. Based on the changes in remaining energy of the EV during each scheduling period, a response capability model considering the charging and discharging process and energy boundary constraints is established. Furthermore, by combining the EV's grid connection time, off-grid time, and expected off-grid energy requirement, constraints are imposed on the range of time periods and adjustable energy that the EV can participate in regulation. The EV response capability model is as follows: ; ; in, For electric vehicles in the first Remaining electricity for each scheduling period For electric vehicles in the first Remaining electricity for each scheduling period For the charging efficiency of electric vehicles. For the discharge efficiency of electric vehicles. For electric vehicles in the first Charging power during each scheduling period For electric vehicles in the first Discharge power during each scheduling period, The time interval between two adjacent scheduling moments. For electric vehicles in the first The lower bound of the energy boundary for each scheduling period For electric vehicles in the first The upper bound of the energy boundary for each scheduling period For electric vehicles in the first The maximum charging power that can actually be achieved during a single scheduling period. For electric vehicles in the first The maximum discharge power that can actually be achieved during a scheduling period. For the grid connection time of electric vehicles, For the off-grid time of electric vehicles, To minimize the function, This represents the maximum energy output of an electric vehicle. This represents the minimum energy required for an electric vehicle. This is the initial charge of the electric vehicle. The expected charge level of the electric vehicle when it is off-grid; Based on the above electric vehicle response capability model, the maximum charging and discharging power and energy boundary of the electric vehicle in each scheduling period can be obtained. That is, the maximum charging power and maximum discharging power that the electric vehicle can actually achieve in each scheduling period, as well as the upper and lower bounds of the energy boundary of the electric vehicle in each scheduling period, can be obtained, thus characterizing the basic response capability of the electric vehicle.

[0025] The air conditioning load response capability model is used to characterize the changes in the average operating power of air conditioners during different scheduling periods within a scheduling cycle. For air conditioning load, an equivalent thermal parameter model is often used to model the changes in indoor and outdoor temperatures. Taking air conditioning cooling as an example, a first-order equivalent thermal parameter model is used to characterize the relationship between indoor and outdoor temperatures and the on / off state of the air conditioner: ; ; ; ; in, This is the equivalent gas specific heat capacity. For the first Indoor temperature during each scheduling period, For the first Outdoor temperature during each scheduling period The equivalent thermal resistance of the building. For the first The on / off status of the air conditioner during each scheduling period. For the first The on / off status of the air conditioner during each scheduling period. For the first Cooling capacity of air conditioners during each scheduling period For the first Heat dissipation from people and various machines indoors during each scheduling period. The energy efficiency ratio of an air conditioner. This refers to the rated power of the air conditioner. This is the lower limit of the air conditioning temperature. This is the upper limit of the air conditioning temperature. Set the temperature for the air conditioner. This is the temperature regulation dead zone, typically set to 2℃, meaning the room temperature is allowed to deviate by 1℃ above or below the set value. When the air conditioner is on and the indoor temperature remains constant, When the value of is 0, the average operating power of the air conditioner is: ; ; in, For the first The average operating power of the air conditioners during each scheduling period Let be the coefficient of the first term of the average operating power function of the air conditioner under thermal steady state. For the first The air conditioning temperature is set for each scheduling period. For the first The adjustment amount of the air conditioning temperature is set for each scheduling period; at the same time, the average operating power of the air conditioning after participating in the response must meet the following constraints: ; in, For the first The maximum upward adjustment that users can tolerate for changes in indoor temperature within a given scheduling period. for For the first The maximum downward adjustment that users can tolerate for changes in indoor temperature within a given scheduling period.

[0026] Based on the above air conditioning load response capability model, the adjustable power range and / or adjustable power range of the air conditioning load in each scheduling period can be obtained, thereby characterizing the basic response capability of the air conditioning load.

[0027] The data center response capability model is used to characterize the adjustment of server computing power consumption and cooling system power consumption during each scheduling period within the scheduling cycle under different load scheduling strategies; data center response capability includes the adjustment of server computing power consumption and cooling system power consumption. ; in, For the first Data center response capability during each scheduling period For the first The adjustment amount of server power consumption for each scheduling period. For the first The adjustment amount of cooling system power consumption during each scheduling period For the first The scheduling period implements the first Total server power consumption after implementing various load scheduling strategies Assign a load scheduling policy number, Indicates spatial adjustment, Indicates time adjustment. For the first Total server power consumption before adjustment during a scheduling period; For the server side, a model was constructed to model the relationship between the total power consumption of the server before adjustment and the number of servers and CPU utilization: ; in, The number of servers in the data center. This refers to the server's idle power. For the first CPU utilization before adjustment during a scheduling period This refers to the server's maximum power. Data centers employ two load balancing strategies: spatial scaling and time scaling. Spatial scaling, also known as interactive load migration, moves loads from one physical location to another data center in real time. Time scaling, also known as batch load deferral, refers to delaying the execution time of loads to a later period without moving their geographical location.

[0028] The total power consumption of the server after space adjustment is: ; in, For the first Total server power consumption after spatial adjustment during each scheduling period For the first The number of servers after spatial adjustment during each scheduling period. For the first CPU utilization after spatial adjustment during each scheduling period For the first Load migration rate for each scheduling period; The total power consumption of the server after time adjustment is: ; in, For the first Total server power consumption after time adjustment during each scheduling period. For the first CPU utilization after time adjustment during each scheduling period For the first Delay rate for each scheduling period; The power consumption of a data center's cooling system is determined by the cooling demand: ; in, For the first Cooling requirements of the data center during each scheduling period For the first The power consumption of the data center's cooling system during each scheduling period. For the first Data center cooling system air supply volume during each scheduling period air density, The specific heat capacity of air at constant pressure. For the first The average temperature of the server cluster during each scheduling period For the first The set temperature of the air outlet of the refrigeration equipment during each scheduling period. This was obtained through a thermal coupling model. and These are empirical constants for refrigeration equipment, generally ; ; The thermal coupling model is as follows: ; in, For the first Total server power consumption during each scheduling period For the first CPU utilization during each scheduling period , Here is an empirical constant related to CPU leakage current, where The value is 0.04 W / ℃ 2 ~0.12W / ℃ 2 , The value ranges from 0.2 W / ℃ to 1.5 W / ℃; This is an empirical constant related to the CPU power supply voltage, and its value is generally taken as 40W~200W; It is a constant, depending on the idle power. Generally, the power is 0W to 100W.

[0029] The adjustment amount for the power consumption of the cooling system is: ; in, To adjust the average temperature of the server cluster, To adjust the average temperature of the previous server cluster.

[0030] Based on the above data center response capability model, the adjustment amount of server computing power consumption under different load scheduling strategies can be obtained. and the adjustment amount of power consumption of the refrigeration system This characterizes the basic responsiveness of data center resources.

[0031] The 5G base station response capability model is used to characterize the dynamic power adjustment and available backup energy storage power of the 5G base station during each scheduling period within the scheduling cycle, as well as the changes in charging power, discharging power, and remaining energy of the energy storage in the 5G base station during each scheduling period within the scheduling cycle. The 5G base station response capability includes the dynamic power consumption adjustment and available backup energy storage energy during operation. ; in, For the first 5G base station response capability during each scheduling period For the first Dynamic power consumption adjustment during each scheduling period's operating state. For the first The dispatchable power of backup energy storage during each dispatch period; Dynamic power consumption adjustment is used to describe the impact of communication load rate on base station power consumption. The dynamic power consumption adjustment is as follows: ; in, , The parameters are linear functions, representing the power consumption in a single 5G base station. Take 1.5kW, Take 2kW; For the first Communication load rate of 5G base stations during each scheduling period; For the backup energy storage portion, an energy storage response capability model is established that considers state of charge, upper limit of charging and discharging power, capacity boundary, and charging and discharging efficiency: ; in, Energy storage in 5G base stations in the first Remaining electricity for each scheduling period Energy storage in 5G base stations in the first Remaining electricity for each scheduling period The charging efficiency of energy storage in 5G base stations. The discharge efficiency of energy storage in 5G base stations. In the 5G base station Charging power during each scheduling period In the 5G base station Discharge power during each scheduling period, This represents the lower limit of energy storage capacity in 5G base stations. This represents the upper limit of energy storage capacity in 5G base stations. For energy storage systems in 5G base stations in the first Reserve energy for each scheduling period Energy storage in 5G base stations in the first Rated charging power for each scheduling period, Energy storage in 5G base stations in the first Rated discharge power for each scheduling period, The energy stored in the 5G base station at the start of the scheduling cycle; This refers to the remaining energy stored in a 5G base station after the scheduling cycle ends.

[0032] Energy storage systems in 5G base stations are in the first The reserve energy during each scheduling period is affected by communication load shifting: ; in, This refers to the power of the active antenna unit (AAU). The power of the 5G base band unit (BBU) For the duration of backup power supply for energy storage; Based on the above 5G base station response capability model, the adjustable power range of the 5G base station in each scheduling period is determined by the dynamic power consumption adjustment amount under working conditions and the charging and discharging power of the backup energy storage, thereby characterizing the basic response capability of the 5G base station.

[0033] The distributed energy storage response capability model characterizes the changes in charging power, discharging power, and remaining energy during different scheduling periods within a scheduling cycle. The response capability of distributed energy storage is primarily constrained by state of charge (SBC), capacity limits, upper and lower limits of charging and discharging power, and charging and discharging efficiency. Based on the energy balance relationship of energy storage in different time periods, a distributed energy storage response capability model considering SBC evolution and charging / discharging boundary constraints is established. This model, combined with the remaining energy requirements at the beginning and end of the scheduling cycle, constrains the sustainable response capability of the energy storage system. The distributed energy storage response capability model is as follows: ; in, For distributed energy storage Remaining electricity for each scheduling period For distributed energy storage Remaining electricity for each scheduling period To improve the charging efficiency of distributed energy storage. For the discharge efficiency of distributed energy storage, For distributed energy storage in the first Charging power during each scheduling period For distributed energy storage in the first Discharge power during each scheduling period, This represents the lower limit of distributed energy storage capacity. This represents the upper limit of distributed energy storage capacity. For distributed energy storage in the first Rated charging power for each scheduling period, For distributed energy storage in the first Rated discharge power for each scheduling period, This represents the amount of electricity at the start of the scheduling cycle for distributed energy storage. This refers to the remaining electricity of distributed energy storage after the scheduling cycle ends; Based on the above distributed energy storage response capability model, the maximum charging power, maximum discharging power and corresponding energy boundary of distributed energy storage in each scheduling period can be obtained, thereby characterizing the basic response capability of distributed energy storage.

[0034] Response capability metrics include the electric vehicle's response time in the first quarter of its lifespan. The maximum charging power that can actually be achieved during a single scheduling period Electric vehicles in the The maximum discharge power that can actually be achieved during a scheduling period , No. Average operating power of air conditioners during each scheduling period , No. Data center response capability during each scheduling period , No. 5G base station response capability during each scheduling period Distributed energy storage in the first Charging power during each scheduling period Distributed energy storage in the first Discharge power during each scheduling period .

[0035] In some embodiments, the factors affecting response capability include those affecting electric vehicle response capability, air conditioning load response capability, data center response capability, 5G base station response capability, and distributed energy storage response capability; correspondingly, the probability distribution models include the electric vehicle probability distribution model, air conditioning load probability distribution model, data center probability distribution model, 5G base station probability distribution model, and distributed energy storage probability distribution model. Factors affecting the responsiveness of electric vehicles include the time of electric vehicle registration and the time of electric vehicle deregistration. Due to the differences in travel patterns among different users, the time of electric vehicle registration and the time of electric vehicle deregistration usually exhibit a multi-peak distribution. Therefore, the time of electric vehicle registration and the time of electric vehicle deregistration are regarded as two-dimensional random variables, and their joint probability distribution is modeled using a Gaussian mixture model (GMM).

[0036] Assuming historical data includes A sample of electric vehicle charging behavior, treating the electric vehicle grid connection time and electric vehicle grid disconnection time as two-dimensional random variables. ,in, Let this be the time vector for electric vehicles entering the grid. For the electric vehicle off-grid time vector, superscript Indicates transpose. Let the time of electric vehicle registration be a random variable. Let the off-grid time of electric vehicles be a random variable. , The time of electric vehicle registration for the first sample of electric vehicle charging behavior. The second sample of electric vehicle charging behavior includes the time of electric vehicle access to the grid. For the first The time of electric vehicle registration in a sample of electric vehicle charging behavior. , The off-grid time of the electric vehicle for the first sample of electric vehicle charging behavior. The off-grid time of the electric vehicle for the second sample of electric vehicle charging behavior. For the first The off-grid time of electric vehicles for a sample of electric vehicle charging behavior; the probability distribution model of electric vehicles is: ; in, for and The joint probability density function, for Possible values, for Possible values, The number of Gaussian components. For the first The weights of the Gaussian components, This represents the Gaussian component index. For the first The probability density function of a Gaussian component distribution. For the first The mean vector of Gaussian components, For the first The covariance matrix of Gaussian components, for The first component, for The second component, for The component in the first row and first column, for The component in the 1st row and 2nd column, for The component in the 2nd row and 1st column. for The component in the 2nd row and 2nd column; ; in, This represents an exponential function with base e. Pi Let be the determinant of the covariance matrix of the m-th Gaussian classification. When a given electric vehicle is registered on the grid At that time, the conditional probability distribution of the off-grid time of electric vehicles is: ; ; ; ; in, The weights are the conditional probability density function for the off-grid time of electric vehicles. For the first The mean of the conditional probability density function for electric vehicle off-grid time corresponding to each Gaussian component. For the first The covariance matrix of the conditional probability density function of electric vehicle off-grid time corresponding to each Gaussian component. For the first The weights of the Gaussian components, To sum the index, For the first The probability density function of a Gaussian component distribution. for The second component, For the first The mean vector of Gaussian components, for The component in the 2nd row and 2nd column. For the first The covariance matrix of Gaussian components; Based on the electric vehicle probability distribution model, the marginal distribution of electric vehicle network access time and the conditional probability distribution of electric vehicle network exit time under a given electric vehicle network access time can be obtained, thereby characterizing the uncertainty of the response period and sustainable response duration of electric vehicles, and providing input for the subsequent generation of initial sample scenario sets.

[0037] Factors affecting air conditioner load response capability include whether the user turns on the air conditioner, the indoor temperature set by the user, and the air conditioner's temperature hysteresis control range. Among these, the air conditioner only has adjustment capability when it is running. Secondly, the indoor temperature set by the user determines the air conditioner's running time. At the same time, the air conditioner's temperature hysteresis control range represents the maximum difference between the actual temperature and the set temperature. The degree of comfort sacrificed by the user affects the air conditioner's adjustment capability, which can be represented by the user's maximum tolerable temperature during the upper adjustment period and the user's minimum tolerable temperature during the lower adjustment period.

[0038] The air conditioning temperature hysteresis control range is: ; in, This represents the actual indoor temperature. This is the lowest acceptable indoor temperature for users during the adjustment period. This represents the highest acceptable indoor temperature for the user during the adjustment period. The indoor temperature set by the user, The lower side temperature hysteresis width, The upper temperature hysteresis width; The user-set indoor temperature conforms to a uniform distribution, that is... , Indicates uniform distribution. The upper limit of the indoor temperature set by the user. The upper limit of indoor temperature set for users; the probability distribution model of air conditioning load is as follows: ; in, The probability density function of the indoor temperature set by the user.

[0039] Factors affecting data center responsiveness include workload; workload is the primary factor influencing data center responsiveness. Workload affects CPU utilization, which in turn affects server power consumption and cooling system power consumption; therefore, its uncertainty directly impacts data center responsiveness. Workload typically exhibits a distinct multi-peak characteristic, such as nighttime troughs, daytime peaks, midday troughs, and evening peaks. Furthermore, this multi-peak distribution is asymmetrical. Therefore, a Gaussian mixture model is used to probabilistically model the workload of the data center during different scheduling periods. Historical workload data for different time periods is obtained from the data center's monitoring system to form a workload time series. The historical workload data is then grouped based on scheduling time periods to obtain a sample set for each sub-time period within the historical period. .

[0040] in, For the first Sample sets for each sub-time period, This refers to the sub-time period number within a historical time period. For the first The value of the first workload in each sub-period. For the first The value of the second workload in each sub-period. For the first Sub-period The value of the workload. The total workload, that is For the first The total number of samples in the sample set for each sub-time period.

[0041] For each sub-period, a probability density function for workload is established to characterize the uncertainty of data center response capability under different typical load conditions, providing input for subsequent scenario generation. The data center probability distribution model is as follows: ; in, for The probability density function, For random variables, for The sample, The number of Gaussian components, i.e., the number of components that may exist in the time period corresponding to this historical time period. A typical load condition, This represents the Gaussian component index. For the first The weights of the Gaussian components, For the first The mean of the Gaussian components, For the first The variance of each Gaussian component.

[0042] Factors affecting the responsiveness of 5G base stations include user behavior factors. The adjustment capability of 5G base stations is affected by user behavior factors, including user population density, the ratio of activated mobile phones to non-mobile video services, the downlink speed of mobile video services, the downlink speed of non-mobile video services, the uplink speed of mobile video services, and the uplink speed of non-mobile video services.

[0043] The ratio of activated mobile phones and the ratio of users using mobile video services to those not using mobile video services are two factors that follow a normal distribution. The probability density functions for the two factors are as follows: ; ; in, for The probability density function, The ratio of the number of activated mobile phones by users. This represents the average ratio of the number of activated mobile phones among users. The variance of the ratio of the number of activated mobile phones. for The probability density function, Comparing mobile video services with non-mobile video services for users, The average ratio of mobile video services to non-mobile video services for users. The variance of the ratio of mobile video services to non-mobile video services for users; The probability density functions for the downlink speed of the user's mobile video service, the downlink speed of the user's non-mobile video service, the uplink speed of the user's mobile video service, and the uplink speed of the user's non-mobile video service are as follows: ; ; ; ; in, for The probability density function, for The probability density function, for The probability density function, for The probability density function, For the downlink speed of users' mobile video services, For the uplink speed of users' mobile video services, For users' downlink traffic speeds for non-mobile video services, For users' non-mobile video service uplink speed; In actual operation of 5G base stations, considering the correlation between multiple influencing factors—for example, periods with high user activation rates typically also see a high proportion of video services—a Gaussian Copula function is used to construct a multivariate joint distribution model to characterize the sources of uncertainty in the 5G base station's response capability and provide input for subsequent scenario generation. The 5G base station probability distribution model is as follows: ; in, for Gaussian Copula joint distribution function, This represents the cumulative distribution function value of the first influencing factor. for inverse function, This represents the cumulative distribution function value of the second influencing factor. for inverse function, For the first The cumulative distribution function values ​​of the influencing factors, for inverse function, This refers to the total number of influencing factors included in user behavior factors. It is the standard normal joint cumulative distribution function; ; in, For the first The cumulative distribution function values ​​of the influencing factors, The influencing factors are numbered. For the first The cumulative distribution function of the influencing factors, For the first One influencing factor, taking the mentioned influencing factor as an example. , .

[0044] Factors influencing the response capability of distributed energy storage include users' willingness to accept regulation. Besides being affected by physical constraints such as state of charge and power boundaries, the actual response capability of distributed energy storage is also influenced by users' willingness to accept regulation. Different users have varying degrees of acceptance of regulation participation, and this difference is typically related to factors such as the energy storage's state of charge, adjustability margin, regulation duration, and user preferences. Therefore, a logistic regression model is used to establish the mapping relationship between various influencing factors and user acceptance of regulation, obtaining the probability distribution of distributed energy storage users' acceptance of regulation under different operating conditions: ; in, For the first Hidden variables for each user Number the user. As the baseline probability, Number the factors influencing users' acceptance of participating in regulation. This represents the total number of factors influencing users' acceptance of participating in regulation. To determine the coefficients of the explanatory variables, the maximum likelihood estimation method was used to solve for the parameters based on the data of each influencing factor in the historical control sample and the actual control results received by users. To affect the first The first user's One influencing factor; The probability of a distributed energy storage user accepting dispatch is related to the functional relationship between various factors as follows: ; in, For the first The probability that a user will accept the scheduling.

[0045] See Figure 3 In some embodiments, Latin hypercube sampling is employed to obtain a set of influencing factors for multi-resource power systems based on probability distribution models. That is, each probability distribution model is used as input to the Latin hypercube sampling method to generate a set of influencing factors for multi-resource power systems. Specifically: Get the initial total number of samples The initial total number of samples represents the total number of initial sample scenarios in the initial sample scenario set, which is the sampling size of the Latin hypercube sampling method. The cumulative probability distribution function is obtained for each probability distribution model; the cumulative probability distribution function is: ; in, For the first The cumulative probability distribution function value of the influencing factors of an adjustable resource. This refers to the numbering of factors influencing adjustable resources. For the first The cumulative probability distribution function of the influencing factors of adjustable resources. For the first Input random variables of factors affecting adjustable resources; Based on the initial total sample size, the value range of each cumulative probability distribution function is evenly divided to obtain a set of value sub-intervals corresponding one-to-one with the cumulative probability distribution function. This is achieved by plotting the cumulative probability distribution function curve in a rectangular coordinate system. Using as the ordinate, Using the x-axis as the horizontal axis, the vertical axis values ​​between 0 and 1 are divided into equally spaced strata, with the total number of strata equal to the initial total number of samples. The system comprises three equally spaced but non-overlapping sub-intervals, with the ordinate interval of each sub-interval being [missing information]. ,in, For the first The lower limit of a sub-interval of values. For the first The upper limit of a sub-interval of values, The sub-interval index is the value sub-interval index within the set of value sub-intervals, and the height of each value sub-interval is . .

[0046] For each sub-interval of values, obtain the initial sample value. Substituting the initial sampled values ​​into the inverse function of the corresponding cumulative probability distribution function yields the sampled values ​​of the factors influencing the corresponding response capability; optionally, when the median value of the vertical axis of each sub-interval is taken as the initial sampled value, the sampled values ​​are: ; in, For the first The first influencing factor of adjustable resources Sampled values ​​of each sub-interval, For the first The median of each sub-interval, that is These are the initial sampled values; Based on a probability distribution model, the sampled values ​​corresponding to each value sub-interval are permuted and combined to obtain the influencing factor scenario set. Specifically: The total number of factors influencing adjustable resources is obtained based on a probability distribution model. ; The sampled values ​​corresponding to the influencing factors of each adjustable resource are arranged and combined to obtain the influencing factor scenario set. Each influencing factor of the adjustable resource corresponds to... The sampled values ​​are obtained by permutation and combination. The dimensional sample matrix represents the set of influencing factors and scenarios.

[0047] In some embodiments, K-means clustering is used to cluster the initial sample scene set to obtain a response capability sample scene set. Specifically: The elbow rule is used to obtain the total number of clusters; specifically: Obtain the cost function, which is: ; in, For the number of clusters is The cost function at that time, Let be the criterion function for the sum of squared errors of the initial sample scene set; Plot the cost function as a function of the number of clusters; the curve is a broken line. The number of clusters corresponding to the inflection point of the curve is taken as the total number of clusters. The inflection point of the curve is the "elbow" position of the curve. The cost function at the inflection point of the curve does not decrease significantly with the increase of the number of cluster centers. The number of clusters corresponding to the inflection point of the curve is the optimal number of clusters, which is the total number of clusters.

[0048] Using the sum of squared sample errors as the criterion function, the error function is: ; in, For the first Error function of the next iteration For the number of iterations, Number the cluster set. The total number of cluster sets, For the first The second iteration The sample scene number in each cluster set. For the first The second iteration The total number of sample scenarios in each cluster set. For the first In the cluster set, the first A sample scene vector, For the first The mean of sample scenarios in each cluster.

[0049] Based on the total number of clusters and the error function, the K-means clustering method is used to cluster the initial sample scene set to obtain the response capability sample scene set.

[0050] See Figure 2 In some embodiments, based on the total number of clusters and the error function, the K-means clustering method is used to cluster the initial sample scene set to obtain the response capability sample scene set, including: A total of 10 initial cluster sets are extracted from the initial sample scenario set; The initial cluster set is used as the current cluster set, and the geometric center of each sample in the initial cluster set is used as the current cluster center. Calculate the distance from each initial sample scene in the initial sample scene set to each current cluster center. Optionally, use Euclidean squares to calculate the distance from each initial sample scene in the initial sample scene set to each current cluster center: ; in, For the first The initial sample scenario and the first During the nth iteration The squared Euclidean distance between the cluster centers; For the first The values ​​of the initial sample scene, This refers to the initial sample scene index in the initial sample scene set. , The initial number of sample scenes in the initial sample scene set, that is... The initial total number of samples, For the first The second iteration The values ​​of the cluster centers corresponding to each cluster set; The initial sample scene set is divided into a total of clusters based on the shortest distance, and the cluster sets are updated accordingly. Determine whether the error function has converged based on the updated cluster set and the current cluster set; If the error function does not converge, i.e. ,in, For the first Error function of the next iteration For the first Error function of the next iteration The preset threshold is typically set to a value of [value to be filled in]. Then, the updated cluster set will be used as the current cluster set, the geometric center of each sample in the updated cluster set will be used as the current cluster center, and the distance from each initial sample scene in the initial sample scene set to each current cluster center will be calculated repeatedly. If the error function converges, i.e. If so, the updated cluster set will be used as the response capability sample scenario set.

[0051] Secondly, this application proposes an adjustable resource response capability assessment method, which uses a probabilistic response capability index to assess the adjustable resource response capability of a multi-resource power system. The probabilistic response capability index is obtained using a probabilistic characterization method for adjustable resource response capability.

[0052] In some embodiments, a probabilistic response capability index is used to evaluate the adjustable resource response capability of a multi-resource power system during each scheduling period within a scheduling cycle.

[0053] In some embodiments, a comprehensive response capability index is used to evaluate the overall response potential of adjustable resources in a multi-resource power system during a dispatch cycle. The comprehensive response capability index is obtained using a probabilistic characterization method for the response capability of adjustable resources.

[0054] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus (systems), or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0055] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce an instruction that executes via the processor of the computer or other programmable data processing apparatus to create an instruction for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0056] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0057] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0058] Finally, it should be noted that the above content is only used to illustrate the technical solution of this application, and is not intended to limit the scope of protection of this application. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of this application shall not depart from the substance and scope of the technical solution of this application.

Claims

1. A probabilistic characterization method for the responsiveness of adjustable resources, applied to a multi-resource power system, wherein the multi-resource power system includes at least two adjustable resources, characterized in that, The method includes: constructing a response capability model for each type of adjustable resource; and obtaining the influencing factors of the response capability for each type of adjustable resource. Based on the factors influencing the response capability, a probability distribution model is constructed for each of the adjustable resources. Using a sampling method, a set of influencing factors for the multi-resource power system is obtained based on the probability distribution model. Based on the influencing factor scenario set, the response capability model is solved to obtain the initial sample scenario set and response capability index value set within the scheduling cycle; Clustering is performed on the initial sample scene set to obtain a response capability sample scene set; A scenario probability set is obtained based on the initial sample scenario set and the response capability sample scenario set; Based on the scenario probability set, the response capability sample scenario set, and the response capability index value set, a probabilistic response capability index is obtained for each scheduling period within the scheduling cycle.

2. The probabilistic characterization method for adjustable resource response capability according to claim 1, characterized in that, The initial sample scene set is clustered using the K-means clustering method to obtain the response capability sample scene set.

3. The probabilistic characterization method for adjustable resource response capability according to claim 2, characterized in that, The initial sample scene set is clustered using the K-means clustering method to obtain a response capability sample scene set, including: The elbow rule is used to obtain the total number of clusters. The sum of squared errors of the initial sample scene set is used as the error function; Based on the total number of clusters and the error function, the initial sample scene set is clustered using the K-means clustering method to obtain the response capability sample scene set.

4. The probabilistic characterization method for adjustable resource response capability according to claim 3, characterized in that, Based on the total number of clusters and the error function, the initial sample scene set is clustered using the K-means clustering method to obtain a response capability sample scene set, including: Extract the total number of initial cluster sets from the initial sample scene set; The initial cluster set is used as the current cluster set, and the geometric center of each sample in the initial cluster set is used as the current cluster center. Calculate the distance from each initial sample scene in the initial sample scene set to each current cluster center; The initial sample scene set is divided into the total number of clusters and the updated cluster set is determined based on the shortest distance. Based on the updated cluster set and the current cluster set, determine whether the error function has converged; If the error function does not converge, then the updated cluster set is used as the current cluster set, the geometric center of each sample in the updated cluster set is used as the current cluster center, and the distance from each initial sample scene in the initial sample scene set to each current cluster center is calculated repeatedly. If the error function converges, the updated cluster set is used as the response capability sample scenario set.

5. The probabilistic characterization method for adjustable resource response capability according to claim 1, characterized in that, The at least two adjustable resources include electric vehicles, air conditioning loads, data centers, 5G base stations, and distributed energy storage; The response capability models include electric vehicle response capability model, air conditioning load response capability model, data center response capability model, 5G base station response capability model, and distributed energy storage response capability model. The electric vehicle response capability model is used to characterize the changes in charging power, discharging power and remaining power of electric vehicles during each scheduling period within the scheduling cycle. The air conditioning load response capability model is used to characterize the changes in the average operating power of air conditioners during each scheduling period within the scheduling cycle. The data center response capability model is used to characterize the adjustment of server computing power consumption and cooling system power consumption during each scheduling period within the scheduling cycle under different load scheduling strategies. The 5G base station response capability model is used to characterize the dynamic power adjustment amount and the dispatchable power of the backup energy storage of the 5G base station in each scheduling period within the scheduling cycle, as well as the changes in the charging power, discharging power and remaining power of the energy storage in the 5G base station in each scheduling period within the scheduling cycle. The distributed energy storage response capability model is used to characterize the relationship between the charging power, discharging power and remaining power of distributed energy storage during each scheduling period within the scheduling cycle.

6. The probabilistic characterization method for adjustable resource response capability according to claim 5, characterized in that, The factors affecting response capability include those affecting electric vehicle response capability, air conditioning load response capability, data center response capability, 5G base station response capability, and distributed energy storage response capability. The factors affecting the responsiveness of electric vehicles include the time when the electric vehicle enters the network and the time when the electric vehicle leaves the network. The factors affecting the air conditioning load response capability include whether the user turns on the air conditioning, the indoor temperature set by the user, and the air conditioning temperature hysteresis control range. The factors affecting the responsiveness of the data center include workload; The factors affecting the 5G base station's response capability include user behavior factors; The factors affecting the response capability of distributed energy storage include users' willingness to accept regulation.

7. The probabilistic characterization method for adjustable resource response capability according to claim 6, characterized in that, The probability distribution models include electric vehicle probability distribution models, air conditioning load probability distribution models, data center probability distribution models, 5G base station probability distribution models, and distributed energy storage probability distribution models. A Gaussian mixture model is used to model the joint probability distribution of the factors affecting the response capability of the electric vehicle to obtain the probability distribution model of the electric vehicle. The air conditioning load probability distribution model is constructed based on the factors affecting the air conditioning load response capability. A Gaussian mixture model was used to model the factors affecting the response capability of the data center to obtain the probability distribution model of the data center; A probability distribution model of the 5G base station was obtained by modeling the influencing factors of the 5G base station's response capability using a Gaussian Copula function. A probability distribution model for distributed energy storage is constructed based on the factors influencing the response capability of the distributed energy storage using a logistic regression model.

8. The probabilistic characterization method for adjustable resource response capability according to claim 7, characterized in that, The Latin hypercube sampling method is used to obtain the influencing factor scenario set of the multi-resource power system based on the probability distribution model.

9. The probabilistic characterization method for adjustable resource response capability according to claim 8, characterized in that, Using Latin hypercube sampling, the influencing factor scenario set of the multi-resource power system is obtained based on the probability distribution model, including: Obtain the initial total number of samples, where the initial total number of samples represents the total number of initial sample scenarios in the initial sample scenario set; Based on each of the probability distribution models, the corresponding cumulative probability distribution function is obtained; Based on the initial total number of samples, each of the value regions of the cumulative probability distribution function is divided equally to obtain a set of value sub-intervals that correspond one-to-one with the cumulative probability distribution function; For each of the aforementioned value sub-intervals, an initial sample value is obtained; Substituting the initial sampled value into the inverse function of the corresponding cumulative probability distribution function, we obtain the sampled values ​​corresponding to the factors influencing the response capability; The influencing factor scenario set is obtained by arranging and combining the sampled values ​​corresponding to each value sub-interval based on the probability distribution model.

10. A method for assessing adjustable resource response capability, characterized in that, The probabilistic response capability index is used to evaluate the adjustable resource response capability of the multi-resource power system. The probabilistic response capability index is obtained using the probabilistic characterization method for adjustable resource response capability as described in any one of claims 1-9.

Images